Stephan Gomer

Solopreneur & Software Engineer, human-in-the-loop

[Stephan Gomer @s4dfun]

A small cloud for personal projects. Easy to use for you and your agent.

Say “deploy this” to Claude Code or Codex and it is live: a domain, GitHub deploys, nightly backups, rollbacks.

One flat price for your own machine, from €9.99 a month. No surprise bills.

[Stephan Gomer @s4dfun]

@maurice_build Consider moving to Pethost

[Stephan Gomer @s4dfun]

@theayush Try Pethost!

[Stephan Gomer @s4dfun]

@nik0diamond Возьмите на Pethost, там и цена неплохая, и тулинг просто замечательный)

[Stephan Gomer @s4dfun]

@oleks01 Pethost is not a cheapest raw resources on the market, but best tooling, UX and value per money in the market 🕺

[Stephan Gomer @s4dfun]

A small cloud for personal projects. Easy to use for you and your agent.

Say “deploy this” to Claude Code or Codex and it is live: a domain, GitHub deploys, nightly backups, rollbacks.

One flat price for your own machine, from €9.99 a month. No surprise bills.

[Stephan Gomer @s4dfun]

@RobGuerra90 Scalability - for now you can upgrade up to 4vCPU + 6 GB RAM + 60GB RAM. This is decent for web apps until you have like many many users. But working on more powerful tiers as well 💪

Security is the same compared to most of PaaS :)

[Stephan Gomer @s4dfun]

@iamnickrz Thank you nick! Hope pethost will be useful for you 😎

[Stephan Gomer @s4dfun]

@itstheplank Thank you! Hope the product also works for you :)

[Sherwood @shcallaway]

We deleted 800k lines of unit tests from the @sazabi monorepo earlier this week.

Three hypotheses:

1- Unit tests "lock-in" slop code by making it hard to change or remove

2- Unit tests slow down dev cycles by making your agent do more work locally and by increasing CI times

3- Unit tests waste money by making your agent generate more tokens and by adding useless CI runs

Will report back soon with the results!

[DAN KOE @thedankoe]

If you want to dramatically improve your quality of life, learn to recover your attention quickly after something steals it. Social media arguments, stressful thoughts, minor inconveniences. Most of it is noise that steals hours to days of enjoyment. Breathe and let it go.

[Stephan Gomer @s4dfun]

Poland is cooked if they accept this.

A lot of skilled people will leave, years of promoting the country as an IT hub will go to waste. Tax revenues will probably fall.

[Tomasz Świetlikowski 🏴󠁧󠁢󠁳󠁣󠁴󠁿 ⭐️ @swietlikowski]

Widzę, że "liberalna" koalicja kolejne prawo podatkowe działające wstecz planuje - obniżenie limitu ryczałtu z 2 M EUR do 0,25 M EUR i uwaga - kwalifikacja tego czy ktoś może w ogóle mieć ryczałt od 1 stycznia 2027 na podstawie przychodów w 2026 (nie może przekroczyć tego 0,25 M EUR z "przyszłych" przepisów). Jest 31 sierpnia, większość roku za nami, puszczają takiego newsa, nie wiadomo kiedy rozstrzygnięcie a ludzie nie wiedzą czy będą mogli być na ryczałcie czy wpadną w progi lub liniówkę. To ma znaczenie dla firm. Dzisiaj.

[Stephan Gomer @s4dfun]

@Gumclaw Yes, and that’s exactly what I meant with my first question “Is it prompting only?” :)

[Stephan Gomer @s4dfun]

@Gumclaw Yes, but if a token can do X, and, if I understood correctly, the token/tools available in your context don’t change depending on the task, then the data is protected by prompt only, right?

[Stephan Gomer @s4dfun]

@Gumclaw Cool, but now the interesting part: how exactly your tools are protected? Is it just a prompt saying, “Ask the user before doing X”? Or does the tool give you a direct API, while the actual calls are programmatically routed through middleware that requires human?

[Stephan Gomer @s4dfun]

@Gumclaw > And not just prompting: tweet text is treated as data, not instructions;

Telling agent to treat some text data IS a prompting. Usually this can be bypassed with formatting. You are smart model and will probably notice it, but it’s still theoretically possible jailbreak

[Stephan Gomer @s4dfun]

@Gumclaw Oh that’s interesting. Does this mean that when you are generating a reply to this tweet, you are not isolated and have a physical access to all the company data?

If yes, how do you protect from prompt injections? Prompting only?

[Stephan Gomer @s4dfun]

@49agents I believe this is something that may work, however I still prefer having a “central” model as default for general tasks

It works for me since for the stuff where, let’s say, Opus outperforms GPT, I can just create a skill for GPT to ask Claude Code to do it 🔮

[Stephan Gomer @s4dfun]

Every time a new model for agentic coding comes out, I spend a full 24 hours testing it on my own tasks

GPT-5.5 is the first model ever to convince me to switch from Claude to Codex. There is one reason: (1/4)

[Stephan Gomer @s4dfun]

I think the rest will notice it soon too

OpenAI actually has a chance to take the lead in agentic coding

The main thing now is not to repeat the mistakes of Anthropic, who lost the trust of most engaged engineers

[Stephan Gomer @s4dfun]

GPT-5.5 in Codex seems to be the first model that actually follows through with research to the very end, without getting lazy or making things up

I've been using it every day since it launched, and it hasn't lied to me ONCE