Show HN: Govbase – Follow a bill from source text to news bias to social posts | Hacker News Reader

Show HN: Govbase – Follow a bill from source text to news bias to social posts

govbase.com

219 pointsfoxfoxx89 comments

Govbase tracks every bill, executive order, and federal regulation from official sources (Congress.gov, Federal Register, White House). An AI pipeline breaks each one down into plain-language summaries and shows who it impacts by demographic group.

It also ties each policy directly to bias-rated news coverage and politician social posts on X, Bluesky, and Truth Social. You can follow a single bill from the official text to how media frames it to what your representatives are saying about it.

Free on web, iOS, and Android.

https://govbase.com

I'd love feedback from the community, especially on the data pipeline or what policy areas/features you feel are missing.

Open articleView on HN

The current summary on the home page contains bias / one-sided reporting.

> While the administration describes the strikes as a necessary move to stop nuclear weapons, the conflict has already seen accidental friendly fire and threats of a ground invasion.

The balance to the assertion "this was necessary" isn't "but there's been some consequences" -- it is an exploration of the truth of the assertion.

And the legality of it as well.

I agree and will be taking this feedback seriously. Daily briefings need more refinement since that is the first thing a user reads.

How are the consequences of war not germane to its necessity?

They are, of course, but there are two different consequences involved in this assessment. One is "stop nuclear weapons" (the converse would be "do not stop nuclear weapons") and the other is "friendly fire incidents" (the converse would be "no friendly fire incidents"). Neither are directly related to the other, since the former is specific to this engagement and the latter happens in any combat.

Yeah, the two sides are basically:

* The war is necessary, regardless of friendly fire

* The friendly fire happened, regardless of whether the war is necessary

They're just totally orthogonal.

That's the same kind of non-balance you see in human-authored news all the time, to be fair.

It also seems rather off base on the sentiment analysis as well.

>"We are on day three of President Trump's military operation in Iran. It's the most courageous military decision of my lifetime, and we are kicking a*. The United States military and the Israeli military, working in tandem, are kicking the hell out of the Iranian government. How is Iran planning to fight back? They have friends. They're counting on pathetic, mewling Europeans and the ridiculous, sad sack Democrats who just hate Trump and don't care about America winning."

This was marked Right:Neutral

Given there is an "AI pipeline" in play, I suspect this is just the typical compulsive equivocation from an LLM. Never assert strong opinions. Find something to say while actually saying nothing. Always give "both sides" equal treatment and consideration no matter what the sides actually are.

That's not a fair assessment. Context: I hate Trump as much as Khomeini. A "both sides" treatment would be:

US & Israel illegally assassinate Iranian leader in bombing campaign, calling agression "necessity".

Now, if you'd like to lean to one side or the other, you can either:

- remove information about legality and the fact that they are the authors of the agression, add something about Iran being a threat to its neighbors

- or insist that any excuses provided by USA or Israel about nuclear weapons is 100% bogus as they have been claiming this for over 20 years

"We have no choice to do this horrible thing, but it may have slightly bad consequences for us" does not take the second side into consideration at all. It's very biased, and it's a very strong opinion in itself.

You are so clearly biased your “both sides” is seething hatred of one side.

Of course i'm biased (though probably not like you mean), but that "both sides" depiction was fair and rather neutral. I'm personally very happy Khomenei is dead, and so are my iranian friends. But we are all very concerned that he is dead for the wrong reason, under a wrong pretext, and with very grim perspectives (see also what the US did in all the countries it bombed in the past 20 years).

I think Khomenei and Trump are two sides of the same coin: bloody authoritarianism and religious zealotry. They're both pretty bad, but one side in this conflict was clearly the aggressor, and denying that is in itself picking sides. One can both sympathize with a victim of unjust aggression, and at the same time thinking they're a profound piece of shit.

One could even point out that just a few years ago, Trump was very insistent about "no more wars", and that he regularly mockingly predicted that Obama would attack Iran to avoid talking about domestic policy. Turns out the hypocrisy level is high and he really is beyond a doubt the bad guy in this story, even if that does not make the iranian ayatollahs good guys by any measure.

"Always give "both sides" equal treatment and consideration no matter what the sides actually are."

It can't even do that correctly. Looking in the list of rights, it has some things called rights and others called policies - "Abortion Rights" vs "Gun Policy". Either call them both policies or call them both rights.

For context, this is a solo project I've been building over the past year while working full-time. I've been responding as "we" in the comments since I got used to doing it other places lol

Looking for feedback and advice. I'm an engineer, not a journalist or policy researcher, so a lot of this domain is still new to me despite working on it for a year.

Love the idea. Thoughts from a UI/UX point of view, on mobile:

* Focus on the policy stuff since that's your differentiator. Put it front and center, currently it's below the "trending news". Nobody needs another trending news feed. I'd cut it entirely.

* Make your differentiator hyper-obvious at a glance on the front page. Right now your above-the-fold is dominated by a wall of AI generated text. It should include a tagline for your site and visuals that people won't get elsewhere.

* Your UI screams "vibe coded" which does not build confidence. Look to other authoritative sites for visual cues - consider a serif for headlines, make your spacing more thoughtful and consistent, reduce or remove your border radius.

Thanks! I'll look into these UX/UI ideas. As for the news, it's front and center because I want Govbase to be a site/app people regularly visit and policy does move slow. Even when a bill is introduced it can take weeks for the actual text content of the bill to show up on congress.gov. Plus on weekends/recess the government doesn't move.

I am planning to bring out more of the impact highlights from the policies to see what's "trending" or what certain reps are working on but just plans for now.

Why chase engagement? If policy is slow-moving then people can visit weekly. Or make an RSS feed. Unless you're planning to go ad monetized or worse...

There's nuance between wanting to build something people use regularly vs "chasing engagement". Even if he decides to run this as a non-profit, individuals are more likely to donate to something they use regularly and institutions are more likely to fund something with active usage. I would assume that the costs to make all these LLM-API costs are not insignificant. I agree with the previous comment that the policy is the differentiator though and hopefully there's a way to drive usage without devolving into a just another news aggregator.

That doesn't make sense to me. I don't need to "engage" daily with MSF to feel like my donation is valuable. I can go days or weeks at a time without using or even thinking much about the Internet Archive, Wikipedia, or my local classical radio station. I almost never hear or see anything about my local food bank/drive organization except when I begrudgingly check a local Facebook group and happen to scroll past a post showing off whatever community dinner they just did.

There is no legitimate reason for a project like this to prioritize a general newsfeed, as opposed to a very specific newsfeed focused on legislation, regulation activity, and court cases. I can think of many interesting and useful ways to integrate the news into a government activity tracker. Yet another slopfeed of whatever nonsense is trending in the news, is not one of them.

It's a shame because I love the idea, but I can't say I trust the creator much at all. I guess now with AI it's easy enough for me to go whip up my own.

I think if you found me on LinkedIn you'd find this comment very ironic. I'm trying to perfect "stories" more before I make them the main feed since they're the core of bringing news, policy, and social posts together.

The idea is to have ordinary people read their news with the facts and the impacts. You don't need to engage with MSF or your food bank for them to make an impact. Bringing policy impact to news does require engagement - it requires interest. This tool isn't for policy researchers and that is exactly why it's not the exact thing you described. Every day people need to use it not people starving or in a war zone.

In this case, I'd say only share news where you have some kind of structured regulatory stuff attached to it. Like sure, Trump started a war, that's noteworthy. Follow that up directly with a link out to structured tally of what reps have said, when was the last time a President did something similar, what are the relevant regulations, etc.

Rather than try to compete in the "current events" space you might have more success in more differentiated channels like having people subscribe to issues, subscribe to reps... News should be part of it but you should lead as much as possible with your differentiator. I bet you could sell enterprise level subscriptions for industry-relevant regulatory news.

This is what "stories" are which I agree will be the headline eventually but its still in development.

Hello,i don't mean to be dismissive but i think lawmaking definitely needs less AI and less pretendly-neutral summaries.

It's already a problem when researching a law proposal that dozens of news outlets will just copy-paste a bland summary of arguments from both sides, neither being explored fully.

I would recommend giving direct links to actually partisan information so people can situate the bill's intent and consequences in a broader context.

Where it helps:

- sometimes a bill's proponent is just an industry puppet whose talking points will be repeated in the media, but aren't solid enough to warrant a proper article… unlike opponents criticizing specific (for example deregulation) points

- sometimes, there are strong feelings and arguments on both sides of a bill and it makes sense to view them in their entirety; seeing one's side unhinged logic sometimes reveals more about the bill than the bill's text itself

- "same-side" opposition: sometimes a bill is perceived as "left-wing" or "right-wing" but receives opposition from the same side; for example, the democrat party is very divided on helping the rich vs taxing the rich, while the republican party is (less than 20 years ago) divided between hardcore authoritarian trumpists and libertarians defending civil liberties

All in all, i believe AI is a plight for society. We are only starting to understand the ecological and psychological costs. There are areas where machine learning can be useful (translation), but i strongly believe politics is not one of them. Please don't try to apply it to anything serious. Don't take it from me, take it from James Mickens in a talk where he explains both how ML works and how it related to the field of computer security:

https://www.usenix.org/conference/usenixsecurity18/presentat...

I would recommend a more systemic mapping like Power Scorecard from Demos.org [1] to help ground the mappings to impacts on populations. Ideally, one would be able to swap the system mappings based on the preferred context, as there are different policies that should be looked at in different ways to assure optimal positive impacts with minimal secondary negative impacts. I haven’t seen a perfect way of doing this yet, but this will be directional correct on the pathway to something even better for all of us. Thank you for your labor of love and efforts!

[1] https://power.demos.org/

> An AI pipeline breaks each one down into plain-language summaries and shows who it impacts by demographic group.

Wont this process be inherently biased by itself? Usually attempts (by humans or computers) to "summarize" or frame things in "plain language" will apply a bias since it intentionally omits all the myriad context and legal/societal "gray areas" that will inform one perspective or another.

Very true. We're constantly trying to refine this and eventually plan on hiring policy researchers for a human in the loop but we just don't have the funding for that currently. We are trying to be transparent for how our scoring does work which you can read more about here: https://govbase.com/methodology

The biggest issue we have found, as you have mentioned, is just the larger context. For example (I don't think this is a real example and would need to check), the TikTok purchase deal could be ranked as an overall benefit for gig workers making content since the outlined alternative was a flat out ban hurting their income. So a deal going through, alleviating that alternative of a ban, in a vacuum is good. However, that ignores the larger context of where that option even came from and the surrounding political context around that deal. So we know the system isn't perfect right now and we're constantly trying to optimize to get the larger picture.

As someone who has been working on this space for a while (not affiliated with govbase) this is really hard. Between eliminating the sycophancy that seems baked into LLMs and dealing with generalized hallucinations - it's freaking hard. I spent this weekend trying to figure out how to get my system to stop telling me the SAVE Act would be fine because it doesn't say what the process for if birth certificate doesn't match current id.

No, I haven't found a good solution yet - I'm going down a rabbit hole of basically crawling the entire federal register for referenced legislation and then adding in an adversarial agent to see if that can spot gaps.

I've been working on the data processing side of legal text with https://www.wordstodata.com/

Your work seems more targeted at tracking the real world impact of the bill rather than the changes it makes to the legal code, but a feature on my roadmap is having bill data also be easily linkable to the votes of politicians so you can track the effect politicians have on the legal code per member. Do you plan to build a member tracker on top of this as well? I think it would be super cool to be able to tie news events to a track record of votes by member of congress.

Yes! If you go to the engage tab you can look up some reps and the policies they're sponsoring but I plan on giving impact scores and where certain reps are focusing their time in the future.

Could you expand out the Examples section to print the output? Would give me a better at-a-glance of what I could expect from the library.

I have long thought about doing something like this with news in general, but I like the specific focus on bills.

What I’d planned to do with news might be a good addition to this system though. If you can, collect news articles from different sources and try to highlight what is left out. I’ve noticed that 99% of news bias isn’t factual incorrectness, but intentional omission of details to frame the story in a specific context.

What you've identified, I call the 2x2 attack (since it can be imagined as a 2x2 grid), which is:

- Report on stuff that helps the positions you support

- Report on stuff that hurts the positions you don't support

- Don't report on stuff that hurts the positions you support

- Don't report on stuff that helps the positions you don't support

It's one of the many, many, fake news tactics that allows them to manipulate and present a false picture of reality, without having to tell a lie.

I like this. My strategy to stay sane in US politics is to follow what the government is actually doing and avoid distractions from ragebait influencers or unhinged statements from politicians.

Thanks! The original goal of Govbase is to make US Policy impact easy to understand for citizens. For more government transparency so people know how their representatives are spending their time and who they're actually working for.

This works great for international geo-politcs as well. It’s much more important to watch what leaders do than listen to what they say.

Looks interesting, but trending social only shows X which will lean conservative. Obviously Bluesky/Reddit will lean left but it should presumably show all bias influences? I don't think Truth Social should be included as its such a niche.

Generally looks like a potentially excellent resource for marketing to media platforms.

Edit: I found a Bluesky one but had to scroll down a lot. If that's to do with relative lack of activity it should probably be clearly explained.

Thanks for the feedback! We mainly include Truth social since Trump and a few of his closest administration are active on there and a large goal of Govbase is to follow the story from Trump posting about tariffs, the news reacting, the EO happening, etc.

I would love to include more Bluesky posts, besides it seeming more balanced - it's also free data compared to X. However, most political social posts happen on X. Even AOC, who is the most followed account on Bluesky still I think is more active on X than Bluesky.

Sad news is that bluesky is on a decline while truth social is increasing.

https://www.forbes.com/sites/conormurray/2025/11/07/bluesky-...

I wonder what will happen to trans kids in the coming years. Trans people shape the soul of the country (since the last 3 years)

It doesn't look like truth social has grown much the last year, sits around 10% of bluesky by active user count. Would be interesting to see more detailed metrics about the amount of content and engagement.

Transgender has been a part of humanity forever, just like we see in other areas of nature. Here's some history going back 150+ years: https://translash.org/articles/drawn-to-history-10-trans-tra...

Zine: https://translash.org/zines/transcestors-trailblazers-30-liv...

Some perspective, if you have an ARM CPU, it's thanks to Sophia Wilson. https://en.wikipedia.org/wiki/Sophie_Wilson

And if you have a CPU at all, it's thanks in part to Lynn Conway. https://en.wikipedia.org/wiki/Lynn_Conway

I expect it's worth looking into the ActivityPub Fediverse, or the applicable IndieWeb protocols (e.g. the linkback quartet: refback, trackback, pingback, and Webmention), to measure what people are saying. None of these have global feeds, though, so you have to deal with the conceptual complexity of measurement. (You have to deal with this anyway with Twitter, because of bot activity, but it's easier to deal with "this is an upper bound" than "this estimate may be too high or too low, and we'd have to investigate further in order to find out which".)

Why does Forbes post intentionally misleading charts?

The one for Bluesky goes from 9M DAU -> 3.5M DAU

The one for X goes from 149M DAU -> 128M DAU

Yet, the Y axis of both charts are wildly out of proportion to make it look like they are equivalent, which is also implied by the headline but clearly not true.

Also, neither is zero based, making a 60% drop (bluesky) and a 15% drop (x) look comparable.

"An AI pipeline breaks each one down into plain-language summaries and shows who it impacts by demographic group."

What controls are in place for bias in this step? Most existing plaintext law coverage has bias in it, so how was it trained to convert the law to plaintext without that biased material?

Edit: Also, where it says it hurts or helps a specific issue, it would be helpful to understand why it was rated that way, especially if there was nothing in it's summary specifically stated about that issue.

Amazing start and look forward to see it evolve. Reminds me a lot of MIT's Open Government Information Awareness [0]. But really like the different track that this takes. I really hope to see this become something people go to.

Suggestion that you increase the education of what you're doing and how. For example looking at the Home Energy Freedom Act [1] some direction to more understanding for each of the sections would be great - what is the process for Legislative Progress - how is the Impact Analysis done. I also couldn't quite figure out if there was a narrative that was being pushed by the parties and how that aligns with media. I like the media ratings though.

[0] - https://www.media.mit.edu/publications/open-government-infor... [1] - https://govbase.com/policy/bill-119-hr-4758

I started but could not finish a project I was calling “g(overnm)it blame” - the idea was to track each bill through committee and to the end either a sort of commit history to see which legislator (or at least which committee) added what part of the final bill.

I found it infeasible, but I’m wondering if you saw rich enough data while making this that you think such a project is viable?

Maybe I'm too software-engineer-brained now, but to me it seems like lawmakers should just be using a tool like git directly. The legal code is a codebase, every bill is a PR, the arguments and proposed changes are captured in review comments, and the PR is accepted/rejected on a vote.

Aside from "lawmakers don't/won't understand the tool", why not do it this way?

I think they pretty much do, it’s just not recorded as such in an easily retrievable format.

From what I understand, it depends on the stage. The United States Code certainly tracks any and all amendments, and you can fully trace which member introduced which amendment, when and where it passed, and even verbatim floor debate.

However, the draft stage isn't documented this way. Members negotiate whatever between themselves (well, really their staffers) and this happens over email, in discussions, via Word documents - whatever works.

I guess in the git metaphor, drafts are in flux while being worked on as a commit, and are squashed and then accessible as such squashed commits once initially introduced or whenever they lead to bill amendments. You can't necessarily track down what member was responsible for a specific sentence in an amendment.

The thing is that there is a voting process eventually so it can at least in principle be known which version is proposed by which legislator, or which committee.

The people behind changes aren't actually as attributable as it sounds though because amendment text gets collaborated on, so showerst might propose an amendment with the key parts of LPisGood's wishlist in it, and then the bill itself will die and then various parts will get cherry picked into an omnibus bill in 6 months anyway.

I've see similar stuff in open source a few times. X post a bug report, Y replies with minimized case, X makes a PR, Z makes a few suggestions, Y add some comments, X updates the PR and W merges it.

Curious about the X/social feature, technically. How are you getting the data? Is it via official APIs or scraping

This is actually really nice. Web page feels pretty snappy, way more so than congress.gov. I've learned some interesting things just scrolling for a few minutes, like the "Energy Freedom Act" cutting appliance rebates or the constitutional amendment for a balanced budget (wtf).

Thanks! This actually started as a mobile app. I hope you check that out too!

This looks pretty interesting. How are you linking the related news for each policy item?

Overall I love it, my biggest critisism so far is that the impact score seems overly opinionated and overly broad in some cases. For example, https://govbase.com/policy/fr-2026-03380 lists on the policy page as positively affecting "Snap Food Stamps" which doesn't seem to be relevant although I don't have pro to see the reasoning.

Interesting. I've been just now exploring something related or tangential, but for Finnish governance (why not all countries eventually).

Idea: Pull in all related docs and statements from the law and its process. And analyze its mechanisms across all timeframes, and across all capital stocks including social/human/moral etc capital. (finding: no one understands game theory nor causal models apparently)

Thank you all for the feedback! I have already started to incorporate some design changes based on suggestions and am looking into AI analysis improvements.

I am also launching on Product Hunt tonight 12am PST if anyone sees this. Would love the support!

https://www.producthunt.com/products/govbase

Hey nice site. The Ledge (https://the-ledge.ai) is a site I developed over the past year with more of a focus on bill content. Ultimately I found it useful to use all three LLM (Gemini, Claude, OpenAi) to ingest and dissect bill text en route to producing useful reports. Especially challenging for bills that run 1000+ pages.

super cool to see a similar project! I'll take a look. Yes I use a mix of Gemini and Claude for certain simple categorizing to full impact analysis.

This looks super useful, but why hide your identity? There's nothing on the website about who you are and I notice you're using a new HN account.

Some of the headlines do not make sense, e.g. https://govbase.com/story/pvxDaH9fXqXUj8yu9Plc. But overall I think this is a great idea.

Yes I just noticed this bug today where there is some character limit impacting story headlines. I appreciate the feedback and will be looking into it today.

One would, naively assume, there's some skew that slowly happens between the source text and the social posts.

In reality, the social posts no longer need to do anything but lie about whatever the title might mean.

Right and the goal with Govbase is to show that disconnect in plain view. Hold representatives accountable to their actions against their words having them right next to each other.

Can you prompt the AI to highlight some "hidden/unexpected causes"?

For example, the bill title say fixing hospitals, but it contains some policy changes about housing.

I think that's a good idea to highlight "unexpected impact". Our system gets the whole policy for analysis so if there is something like housing impacts within a medical focused bill, Renters and Home Owners impacts should show up.

If you can figure out how to actually make it sound unbiased on the front page and then route to both sides biased versions that would be something.

Is it federal-only?

Yes but state level actions are definitely on the roadmap!

I subscribed. Awesome idea and execution. Hope I see more apps working to make this stuff more accessible!

>shows who it impacts by demographic group.

Why do we need to bake this zero-sum mentality into everything?

to be clear impact groups are things like government programs (e.g. snap benefits), disabilities, immigration status, etc. They're not "democrat vs republican". The whole point of Govbase is for users to be able to follow things that matter to them and see how the government is impacting their lives.

You can take a look at https://govbase.com/methodology for more info.

What's the recommended way to consume this with other ai? What kind of api is available?

The dismiss button on the top banner doesn't work after I click onto the trial page.

Thanks for the feedback! We've tested it on a few different browsers. Can I just get what browser you're using and if on mobile? Thanks!

Not able to create account via Apple - invalid_request: Invalid web redirect url.

Thank you! The web app launched less than a month ago so definitely still working on bugs. I will get this fixed today. I appreciate the notice.

Should be fixed! If you'd like to try again. Thank you.

If bluesky is included no reason to not include mastodon, threads, instagram,

Understandable. Our priority right now is politician and agency posts though and these accounts are just not very active on mastodon or threads. We can look further into it and another comment mentioned this too.

There are two parts here

1. Platforms politicians, governments, and media

2. Platforms which have an open (and free?) API

Bluesky seems to be the only one covering both, though less coverage on #1 than others, minus Mastodon

Checkout has been fixed for anyone who wanted to try Pro!

We need something similar for the UK

Feels a bit strange to use or at least not what I was expecting. I'm not sure having a "feed" the way it does is even appropriate, but assuming that's what it will be, so be it. The titles, however, read like headlines even under the "Policy" tab, and it isn't until clicking through that I can see the title of the bill in question and some brief description of it. I was expecting something more like a list of bills with outbound links to discussions and press releases where they may be.

I'm not sure the bullet points make a lot of sense on the impacts, either. The first one right now is a bill to change security rules for hospitals and healthcare systems that offer remote logins to retrieve patient details. You only find that's what it is by scrolling all the way to the bottom and finally reading the summary, but first you see a list of impacted parties and it highlights people with chronic illnesses and tribal members. I think I at least understand the logic of the first one, assuming chronically ill log into patient portals more often than healthier people, but it feels somehow facile, like saying a bill about highway maintenance affects drivers more than non-drivers. No shit. That isn't really an insight and shouldn't be above the actual content of the bill.

The "source information" is also all the way at the bottom even though, personally, it's what I would care about the most. And it has no links at all. You can look up the bill number and find it in the congressional database, but why not include a direct link? The news snippets link to the sources they came from. Why not the bills themselves?

So actually, I can see now there is a link to the bill itself. It's just all the way at the bottom and not part of the source summary, whereas the news summaries are tiles that also act as links all on their own. I guess the question is why make that different and why put the link I most care about all the way at the bottom beneath all of the information? Not gonna lie, though. I almost hesitate to ask because I fear the answer is there is no known reason. You asked an AI to put together a page and this is what it did. There is no knowable "why" and even though you're publishing this as if it's your product being created based on your design decisions, it isn't.

So policy feeds are dynamic news headlines because no one really wants to click on or understands:

"Making further consolidated appropriations for the fiscal year ending September 30, 2026, and for other purposes."

This is similar to how most people don't want to read the raw source bill text which is why it's at the bottom. The reason policies aren't direct links to congress.gov is because I've spent more time than most on congress.gov and the federal register and on one wants to do that.

My first commit on this project was Feb 22, 2025 so I'm sure you're happy to find out there is plenty of "why" to my answers and that these are all my design decisions.

Well intentioned, but very naive.

We're early stage, but I believe there's a space between news and actual policy that no one's filling well. If you can show people that their representatives are making their lives better or worse, with real policy behind it, they'll care.

Right now, too many people are consuming misinformation from sources they believe are legitimate, and increasingly from social media where real people are getting their news. We need to connect the policy, the personal impact ("you're losing your insurance because of X"), the news, and what politicians are actually saying, all in one place, to bring real facts to the misinformation and make government more transparent.

I agree but what's your reasoning? Politico pro subs or just generally that no one cares?

Reply on news.ycombinator.com